On first start the application needs to be setup.
The application is using two 'rounds' to find the most optimal overclock settings for each GPU in a system. The first round is for attempting to find which combinations of 'power target', 'base clock offset' and 'memory clock offset' the GPU's will not work with. The second round takes the top results from the first round and runs each at a longer period of time. Round 2 attempts to find if a combination of parameters is stable over time, to try and make each system as stable as possible while running overclocked GPU's. The number of runs in round 2 can be set using the 'Round 2 runs' setting. Different algorithms will require different length's of each run. For example it is highly recommended to use, if not the maximum time for each round, at least 1 hour for round 1 and at least 9 hours for round 2 when evaluating x16r algorithm. Other algorithms require's less time for each round, but the longer the runs are, the more likely it is that the results found by the application will be stable over longer period of times when mining.
If the checkbox 'Use the same parameters for all devices' is checked the user don't have to input parameters for each GPU on the 'Parameters' tab. Instead only the first GPU will be enabled on that tab, and the parameters for the rest of the GPU's will filled in as the user chooses parameters for the first GPU. If the checkbox is not checked the user will have to input parameters for each GPU individually and click each save button for the parameters to be added to the database.
There are currently two different logics available when evaluating.
⦁ Thorough, will walk through all possible combinations beginning with lowest settings. First it will go through base clock offset and for each successful run it will increase the offset by the value set in the 'base clock offset increase steps' box. If the miner crashes during a run it will retry the same paramaters as many times as specified in the 'Round * retries' options. If all attemps fails at a given base clock setting the application will mark that combination as failed in the database, and it will assume that any higher base clock would fail as well and mark them as failed. When it reaches the highest value set for base clock offset, or if the miner fails and the rest is marked as failed, the application increase the memory clock offset by the steps defined in the 'memory clock offset increase steps', and the base clock will start from the lowest defined offset again. For each memory clock offset the application walk through the available core clock offset steps again. When the steps for memory clock offset has reached the it's maximum, the application will start increase the power target and both core clock offset and memory clock offset will start from the lowest defined value.
Depending on the settings used this can be very time consuming.
⦁ Random, works a little bit differently. It will pick a set of parameters from the database and run through it. It will still use the retry amount defined, and it will still mark any higher core clock offsets as failed if the miner crashes. If a run is successful, it will however also assume that any lower core clock offset would also be successful and mark them as such in the database, meaning it will run those mark as successful. This will decrease the length of the evaluations quite a lot, depending on the ranges set in the parameters. The 'estimated time remaining' will be the same when an evaluation is started but will decrease during the runs.
This logic is faster, but it will not be as precise as using thorough. It is however recommended for algorithms that needs longer runs for evaluating if they are stable or not.
As can be seen in the screenshot below there are an option for using a custom fan curve. The application contains two simple curves that gives the user the choiche not to use any other tool for controlling the GPU fans. This option will however not be available for users that have RTX cards in their systems, simply because the tools used to control the fans do not support fan control on those cards at the moment. So a third party tool is recommended.
Skip seconds are used to control how much time a miner should be given to be able to connect to the pool and start mining. This can differ quite a lot between miner starts, so at least 30 seconds will be required for most miners and algorithms. Some will need even more time. This is used so that each run can be measured only with data from actual mining and not from waiting for data. This affects both the hashrate and the power draw.
Intensity increase steps sets how big the steps are to be when evaluating intensity.
Use Tp Link Smart Plug (HS110) enables to get the powerdraw from the plug (wall) instead of just the GPU's. Check the box and enter the ip address for the plug. The 'Total powerdraw' at the top will when show the wall powerdraw. If unchecked it will show the sum of all GPU's in the system. If an incorrect ip address is entered it will show 'ERROR' instead of current powerdraw, it might take a few seconds to update.